Effectiveness validation of the Target-AID base editor inS. cerevisiae
The Target-AID base editor, composed of gRNA and the fusion protein of Cas9 nickase (nCas9, D10A), PmCDA1 (an AID ortholog) and UGI (uracil DNA glycosylase inhibitor protein), was recently developed to enable C-to-T substitutions without DSB and donor DNA in S. cerevisiae [27]. The CAN1 and ADE1 genes were targeted to test the effectiveness of Target-AID by introducing a stop codon via C to T mutation. Here, to further assess whether the Target-AID would be a generally effective tool to create site-specific point mutagenesis in S. cerevisiae, we selected four sites from the URA3 and ADE1 genes as targets, following the reported criteria that the cytosine mutations in the − 20 to -13 position upstream of the PAM may introduce a stop codon (Fig. 1a, b) [27]. The introduction of early stop codons in URA3 and ADE1 can lead to their gene disruption and functional loss, thus resulting in a 5-FOA resistance phenotype and a red colony phenotype, respectively. Furthermore, the iterative strategy commonly used in genome engineering was employed as well [53]. After 5 to 6 generations, the two target sites of URA3 resulted in approximately 8.8% and 7.3% mutation efficiencies (Fig. 1a, b). By contrast, the two target sites of ADE1 resulted in approximately 53.1% and 43.7% mutation efficiencies after 3 to 4 generations (Fig. 1a, b), which were higher than the previously reported mutation efficiencies of 16–47% at other four target sites of ADE1 [27]. Additionally, base-edited cells showed more than 60% cell viability (Fig. 1c), further confirming that Cas9 nickase-based Target-AID might cause less growth defects or cell death than the full CRISPR/Cas9 nuclease-mediated approaches [19]. These results indicated that gene- and site-specific biases were also observed for Target-AID-mediated point mutagenesis. However, compared to site-specific point mutagenesis by CRISPR/Cas9 practices [54–56], the Target-AID system avoids high cell toxicity and some tedious processes, such as design and usage of heterology block or stuffer to prevent repeated recognition and cutting by the gRNA/Cas9 complex, preparation of donor DNAs, etc. Therefore, Target-AID could provide a powerful and effective tool to generate site-specific point mutagenesis in S. cerevisiae.
Computational and experimental scanning mutagenesis ofSPT15via Target-AID base editor
Beyond gene disruption, we aimed to explore the application of the Target-AID in generating nonsynonymous mutations. A general transcription factor Spt15 (TATA-box binding protein), whose point mutations have been partially investigated to influence its interactions with DNA and other proteins in the yeast transcription machinery and/or enhance yeast stress tolerance [35–38], was selected as a target. Spt15 is encoded by a 723-nt nucleotide sequence, and composed of 240 amino acid residues. First, a point mutation of alanine at each position of the entire Spt15 was computationally scanned to analyse its variant impacts on protein functions using the MutFunc database [57]. Approximately two-thirds of 240 amino acid residues were predicted to have single or multiple effects on conserved regions, protein-protein interaction interfaces, phosphorylation, and protein stability (Fig. 2a, Additional file 2: Table S1). Second, the potential Target-AID target sites of the entire Spt15 to introduce nonsynonymous mutations were computationally analyzed. These sites should meet two criteria: (i) cytosines localize in the editing window of the − 20 to -13 position upstream of the PAM on the sense or antisense strands and (ii) codons containing cytosine mutations can lead to nonsynonymous mutations. These potential sites covered 24.6% of total 130 cytosines on the sense strand and 21.4% of total 159 cytosines on the antisense strand (Additional file 3: Table S2), possibly leading to nonsynonymous mutations at 50 positions of the entire Spt15, 60% of which were predicted to have important variant impacts in the MutFunc database (Additional file 2: Table S1). Computational scanning mutagenesis suggested that Spt15 is a feasible and promising target to alter its function by creating nonsynonymous mutations via Target-AID, and thus manipulating yeast stress tolerance.
As an experimental practice, individual site-specific gRNA plasmids and the nCas9(D10A)-PmCDA1 plasmid were employed to in situ mutagenize SPT15 by targeting 33 predesigned PAMs in the genome (Additional file 4: Table S3), and next to screen mutant strains harboring SPT15 point mutations and with enhanced stress tolerance (Fig. 2b). We obtained total 36 Spt15 mutant strains that had nonsynonymous mutations at half of the 50 predesigned positions of the Spt15 by mutagenizing 49 bases (Fig. 3b). Additionally, except for 28 C to T mutations, 19 C to G and 2 C to A mutations were also found (Additional file 3: Table S2). This observation consisted with the previous report [27], and thus resulting in more than one nonsynonymous mutation at the same amino acid position. Among them, 13 mutants at 8 amino acid positions were localized at the N-terminal region (amino acid residues 1–60). The rest mutants were scattered at 10 β-sheets and 4 α-helixes of the C-terminal domain (residues 61–240), which displays a saddle-shaped tertiary structure as previously reported (Fig. 3b, c) [38]. The concave surface of the saddle is dominated by ten antiparallel β-sheets and responsible for DNA binding, while the convex surface of the saddle is governed by two large and two small α-helixes and outwardly available for interaction with other proteins during transcription initiation [58]. Further mapping of these mutations on the tertiary structure of Spt15 showed that 12 mutants at 9 amino acid positions were localized at the concave surface, and 10 mutants at 7 amino acid positions were localized at the convex surface (Fig. 3c). P187A was situated on the stirrup between helix 2’ and 3’ (Fig. 3c). Localization of these mutations in different functional domains implicated distinctive ways in which they might influence the function of Spt15, thus inducing diversified effects on yeast stress tolerance capacities.
Stress tolerance variations of Spt15 mutations
Fermentation experiments were performed to determine stress tolerance capacities of the Target-AID generated 36 Spt15 mutant strains. First, all the fermentation data (cell growth, glucose consumption and ethanol production) at different time points from all the strains at four conditions including normal, hyperosmotic, thermal and ethanol conditions were subjected to principal component analysis (PCA). The results showed a clear separation of samples into condition-dependent clusters (Fig. 4a). In particular, the variation range of the fermentation data across strains at each of the three stress conditions were much larger than that at normal condition, indicating diversified stress tolerance capacities of these Spt15 mutant strains.
To further look into the differences of stress tolerance capacities of these Spt15 mutant strains, the fermentation data at each of the three stress conditions were separately used for PCA (Fig. 4b, Fig. S1a, b). In addition, fold changes of fermentation parameters, such as cell growth (ΔOD600), glucose consumption (ΔGlucose) and ethanol production (ΔEthanol), at a certain time point of log phase in comparisons of mutant strains versus the wild type strain were analyzed as well (Fig. 5). Compared to the wild type strain at each batch experiment, mutant strains localized at the right or upper area on the PCA plots had increased stress tolerances, while strains situated at the left or lower area had decreased stress tolerances. Furthermore, the farther the mutant strains were separated from the wild type stain, the more significant the mutant stains were stress tolerant or sensitive. For instance, the mutant strains, such as E9K, P169A and R6C from the first batch, L214F, T38I, W26S and A101P/V102I from the second batch, R238R, A150P and V102L from the third batch, as well as D56E and P20L from the fourth batch, showed relatively higher stress tolerance capacities in contrast to the wild type strain at hyperosmotic stress condition (Fig. 4b). Additionally, L214F, T38I, W26S and A101P/V102I of the above strains as well as W26C and A140G showed more than 1.5-fold increases of fermentation parameters at the 24-h time point of log phase in contrast to the wild type strain (Fig. 5a). On the other hand, the mutant strains, such as S118L, M197I, L214V and R6G from the first batch, S42N from the second batch, R141K, G174V, T124S and T124I from the third batch, as well as V71I from the fourth batch, showed relatively lower stress tolerance capacities in contrast to the wild type strain at hyperosmotic stress condition (Fig. 4b). Additionally, S118L, L214V, R6G, S42N, G174V and V71I of the above strains showed lower than 0.5-fold decreases of fermentation parameters at the 24-h time point of log phase in contrast to the wild type strain (Fig. 5a).
As for thermal stress condition, the result of PCA distinguished a same set of significantly tolerant and sensitive mutants as hyperosmotic stress condition (Fig. S1a). Furthermore, L214F, T38I, W26S, A101P/V102I, V102L and D56E of the PCA distinguished stress tolerant mutants as well as W26C, A140G and R238K showed more than 2.5-fold increases of fermentation parameters at the 18-h time point of log phase in contrast to the wild type strain, whereas R6G, S42N, S118L, T124I, T124S, R141K, G174V, M197I and L214V of the PCA distinguished stress sensitive mutants as well as A2V and P191C showed lower than 0.5-fold decreases of fermentation parameters (Fig. 5b).
In terms of ethanol stress condition, the PCA result distinguished a slightly different set of significantly tolerant and sensitive mutants compared to hyperosmotic and thermal stress conditions (Fig. S1b). To be detailed, the mutant strains, such as P169A, E9K, R6C and A2D from the first batch, A101P/V102I, L214F and A140G from the second batch, V102L, R238K and R238T from the third batch, as well as P20L from the fourth batch, showed relatively higher stress tolerance capacities in contrast to the wild type strain at hyperosmotic stress condition (Fig. S1b). Additionally, R6C, E9K, P20L, A101P/V102I, V102L, A140G, P169A, L214F, R238K of the above strains as well as W26S showed more than 3-fold increases of fermentation parameters at the 30-h time point of log phase in contrast to the wild type strain (Fig. 5c). On the other hand, the mutant strains, such as L214V, M197I, S118L, A2V, P191C and R6G from the first batch, S42N from the second batch, R141T, R141K, A150T, G174V, T124S and T124I from the third batch, as well as P20R and V71I from the fourth batch, showed relatively lower stress tolerance capacities in contrast to the wild type strain at hyperosmotic stress condition (Fig. S1b). Additionally, A2V, R6G, S42N, S118L, P191C, M197I and L214V of the above strains as well as T38S and P187A showed lower than 0.5-fold decreases of fermentation parameters at the 30-h time point of log phase in contrast to the wild type strain (Fig. 5c).
Besides the above results, the range of the fold changes of fermentation parameters at normal condition were smaller than those at hyperosmotic and thermal conditions (Fig. 5d). Except that P65L, R141K, P191C and L214V showed lower than 0.5-fold decreases of fermentation parameter, the fold changes of fermentation parameters of the rest mutant strains were in a range of 0.5 to 1.5 in contrast to stress conditions (Fig. 5d). This result further confirmed diversified stress tolerance capacities of these Spt15 mutant strains as observed in the PCA result using all the fermentation data (Fig. 4a).
Taken together, whether the 36 Spt15 mutants were stress tolerant or sensitive were qualitatively annotated in Fig. 3b. The number of stress tolerant mutants obtained at the N-terminal region (amino acid residues 1–60) was eight, whereas the number of stress sensitive mutants was three. Five stress tolerant and four sensitive mutants were found to be at the concave surface of the tertiary structure of Spt15. Four stress tolerant and seven sensitive mutants were found to be at the convex surface of Spt15. Thus, it seemed that the obtained stress tolerant strains were enriched at the N-terminal region (amino acid residues 1–60) and the convex surface, while the obtained stress sensitive strains were enriched at the concave surface.
Combining the evaluation results of stress tolerance capacities and variant impact analysis in the MutFunc database (Additional file 2: Table S1), we selected three stress tolerant mutant strains (A140G, P169A, R238K) and two stress sensitive mutant strains (S118L, L214V) for the following transcriptome and protein structural analyses to reveal the underlying mechanisms of regulating yeast stress tolerance by Spt15 mutants. At normal conditions, the A140G, P169A and R238K mutant strains showed similar fermentation performances to the wild type strain, while the S118L and L214V mutant strains exhibited reduced fermentation performance (Fig. 4c). At stress conditions, the A140G, P169A and R238K mutant strains showed better fermentation performances than the wild type strain to different extents, while fermentation capacities of the S118L and L214V mutant strains were severely hampered compared to the wild type strain (Fig. 4c). Thus, these results suggested potentially distinctive regulatory mechanisms underlying different Spt15 mutants.
Genome-wide reprogramming of global transcription regulated by key Spt15 mutations in response to stress
To validate the predicted distinctive regulatory mechanisms underlying different Spt15 mutants (A140G, P169A, R238K, S118L and L214V), genome-wide transcriptome analysis by RNA sequencing was carried out to quantify global transcription changes in the Spt15 mutant strains compared to the wild type strain at the same culture conditions including the unstressed normal condition as well as hyperosmotic, thermal and ethanol stress conditions. To be noted, RNA sequencing of samples grown at ethanol stress conditions and samples from the S118L mutant strain grown at thermal stress conditions failed due to that qualified RNA cannot be extracted from their severely hampered cell growth. Significantly differentially gene expression (SDEG, absolute fold changes ≥ 2.0 or more; FDR p-value ≤ 0.05) in comparisons of the Spt15 mutant strains versus the wild type strain at each culture condition were extracted, revealing differential global transcription changes in these five Spt15 mutant strains (Additional file 5–7: Dataset S1-S3). Regardless of unstressed or stressed conditions, the stress tolerant strains (A140G, P169A, R238K) displayed hundreds of SDEGs, while the stress sensitive strain S118L showed more than one thousand SDEGs (Fig. 6a), which seemed to be coincided with its more significant decreased changes of fermentation capacity in contrast to the wild type stain (Fig. 4c). Although the other stress sensitive strain L214V showed the least number of SDEGs, the percentage of down-regulated SDEGs was much higher than that of up-regulated SDEGs, which could explain its more significant decreased changes of fermentation capacity in contrast to the wild type stain (Fig. 4c). Among SDEGs, four to sixty-three significantly differentially expressed transcription factors (SDE-TFs) were found in different mutant strain comparisons and showed similar patterns in terms of the number and the percentages of up- and down-regulated TFs (Fig. 6a).
The Spt15 mutant strains showed quite different alterations of GO biological process enrichment, which were influenced by different culture conditions as well (Additional file 8: Dataset S4, Fig. S2). It seemed that up-regulated protein translational modification, sulfur assimilation, amino acid biosynthesis and iron ion homeostasis might be beneficial to the acquirement of stress tolerance due to the Spt15 mutations including A140G and P169A. By contrast, the stress sensitive performance of the S118L strain might be due to the energy imbalance caused by down-regulated ATP producing carbohydrate metabolism and up-regulated ATP consuming translation.
As a general transcription factor, Spt15 plays functions by working with many other transcription factors (TFs), such as specific activators and repressors, in preinitiation complex assembly to regulate gene transcription [59]. Amino acid mutations of Spt15 might influence its interplay with specific activators and repressors, thereby altering the global transcription. Thus, we employed the TFRank method in the YEASTRACT database (http://www.yeastract.com) to deduce specific transcription factors and SDE-TFs clustering in SDEGs identified in different Spt15 mutant strains [60, 61]. Clustered TFs and SDE-TFs targeting to more than 50% SDEGs in each comparison were considered (Additional file 9: Dataset S5A and S5B). Remarkably, most of these clustered TFs were found to be well-known stress-responsive TFs, such as Msn2/Msn4, Bas1, Yap1, Pdr3, etc. (Fig. 6b) [8, 62]. Except for the A140G strain, all the other four mutant strains showed that the great majority of clustered TFs at normal conditions were also enriched at hyperosmotic and thermal conditions (Fig. 6b) [8, 62]. As for the A140G strain, much more TFs were clustered with SDEGs at hyperosmotic and thermal stress conditions than at normal conditions (Fig. 6b). These results suggested that the four Spt15 mutants including P169A, R238K, S118L and L214 might intrinsically interplay with those stress-related TFs at normal conditions, thereby dominantly determining the stress tolerant or sensitive capacities of cells against future stresses. However, the interplay between the A140G mutant and newly involved TFs might be induced by applied stresses, implicating a different role of this amino acid position in the Spt15 from the other four positions.
We further compared the differences of the clustered TFs and SDE-TFs between the stress tolerant and sensitive Spt15 mutant strains to reveal transcriptional regulatory hubs impacted by the Spt15 mutations. Msn2/Msn4, Bas1, Tec1 and Ste12 were found to be commonly highly clustered (Fig. 6b). Msn2 and Msn4 regulate general stress responsive gene expression against several stresses including heat shock, osmotic shock, oxidative stress, low pH, glucose starvation, sorbic acid and high ethanol concentrations [63, 64]. Msn2/Msn4 regulated genes encode chaperones, certain antioxidant enzymes, housekeeping enzymes, proteases, and other proteins involved in the removal of damaged biomolecules and restoration of metabolic homeostasis or in repair processes [8]. As previously reported, MSN2 gene expression is constitutive, while MSN4 expression is Msn2/4 dependent and induced by stress [64]. Here, two factors including Spt15 mutants and environmental conditions should be considered. Msn2, clustering with more than 50% SDEGs that were influenced by the mutant A140G and by R238K at normal non-stressed and thermal stress conditions, showed significantly increased transcription expressions compared to the wild type strain (Fig. 6c). By contrast, Msn4, clustering with more than 50% SDEGs that were influenced by the mutant P169A, R238K and S188L at hyperosmotic stress conditions, showed significantly decreased transcription expressions compared to the wild type strain (Fig. 6c). These results suggested that the regulatory effects of Spt15 mutations on Msn2 and Msn4 transcriptions seemed to be specific to normal non-stressed and thermal stress conditions or hyperosmotic stress conditions, respectively. Bas1, regulating some genes in the purine and histidine biosynthesis pathways and in meiotic recombination, is predicted to regulate stress responses under oxidative stresses, temperature shift stresses and osmotic stresses [65]. Bas1 showed significantly increased transcription in the stress tolerant R238K mutant strain but significantly decreased transcriptions in the stress sensitive S118L mutant strain grown at hyperosmotic stress conditions. Tec1 and Ste12 are involved in hyperosmotic stress response and regulate most genes in the filamentation/invasion pathway [66]. Tec1 had no significant transcriptional change, whereas Ste12 showed significantly increased transcription in the S118L strain at normal condition.
Yap1, Sfp1, Pdr3 and Gcn4, which are well-characterized TFs in response to various stresses [63, 64], were commonly highly clustered in regulating SDEGs of all three stress-tolerant mutant strains and the stress sensitive S118L strain, but not the other stress sensitive L214V strain (Fig. 6b). Additionally, Yap1 showed significantly increased transcription in the A140G strain at normal condition as well as in the R238K strain at normal non-stressed, hyperosmotic and thermal stress conditions compared to the wild type strain, but not in the P169A and S118L strains. And no significant transcriptional changes were observed for Sfp1, Pdr3 and Gcn4. Most interestingly, Met28, Met31, Met32 and Cbf1, which are cofactors of the Met4 transcription factor controlling the expression of sulfur metabolism and oxidative stress response genes [67, 68], were commonly highly clustered in regulating SDEGs of the stress tolerant P140G and R238K strains as well as the stress sensitive S118L strain (Fig. 6b). In the comparisons of the P140G and R238K strains versus the wild type strain, Met28 and Met32 showed significantly increased transcription at normal and/or hyperosmotic stress conditions, while Met28 showed significantly decreased transcription at hyperosmotic stress condition in the S118L strain. Additionally, Cbf1 showed significantly decreased transcription in the R238K strain at hyperosmotic stress condition, but significantly increased transcription in the S118L strain at normal condition. Besides the above highly commonly clustered TFs, Rap1, Sok2, Rpn4, Pdr1, Cst6 and Ixr1, which are involved in stress response [8, 62, 69, 70], were also observed to be clustered in regulating SDEGs of both the stress tolerant and sensitive strains, although they had no significant transcriptional changes (Fig. 6b, c). In addition, four other well-known stress-related TFs including Fhl1, Hsf1, Cin5, and Ume6 were specifically clustered in SDEGs of the L214V strain (Fig. 6b) [31, 69, 71].
Predicted impact of key amino acid changes on protein structure and function of Spt15
As a general transcription factor, Spt15 binds to the TATA-box and interacts with other factors to form the preinitiation complex at promoters. We predicted that key amino acid changes might have impacts on the interactions of Spt15 with DNA and other proteins by influencing Spt15 conformation, and thereby resulting in genome-wide reprogramming of global transcription in the Spt15 mutant strains. Thus, to assess potential conformational changes caused by amino acid changes, we performed pairwise protein structure alignment between Spt15 point mutation and the wild type Spt15 in two crystal structures of Spt15 related complexes (PDB 1YTB and 4B0A) (Fig. 7) and calculate root mean square deviation (RMSD) values. 1YTB shows the interaction between Spt15 and the TATA-box DNA [72], whereas 4B0A contains the Spt15 and Taf1 N-terminal domains TAND1 and TAND2 to show competitive multiprotein TBP interplay patterns [73]. Superposition structure of the Spt15-DNA (Chain A and C) from the structure 1YTB and Spt15-Taf1 from the structure 4B0A showed a competitive binding of Taf1 at the concave DNA binding surface (Fig. 7a). The profiles of RMSD versus each position at DNA in 1YTB indicated apparent peaks at positions 11 to 22 of DNA due to the A140G and P169A mutations (Fig. 7b). Correspondingly, significant conformational changes at positions 11 to 22 of DNA, which is localized downstream of the TATA box, were observed in pairwise protein structure alignment between Spt15 point mutations of A140G and P169A and the wild type Spt15 in crystal structure 1YTB, which were relatively smaller in R238K, S118L and L214V mutants (Fig. 7c). Upon DNA opening during transcription initiation, melting begins about 20 base pairs downstream of the TATA box [58, 74]. Thus, Spt15 point mutations of A140G and P169A might impact DNA opening during transcription initiation. The RMSD profiles at amino acid positions of 60 to 240 in Spt15 of 1YTB implicated that these five mutations might have effects on Spt15 conformational changes to different extents (Fig. 7b). The P169A mutation seemed to induce a bigger conformational change than the A140G mutation at the C-terminal stirrup (Fig. 7c). The R238K mutation caused an obvious conformational change between the sheets S3’ and S4’. The S118L mutation at the end of the S4 sheet resulted in an apparent conformational change between the sheets S3 and S4. The L214V mutation induced a slight conformational change between the S5’ sheet and the H2’ helix. These differential conformational changes of Spt15 suggested distinctive impacts of these five mutations on the function of Spt15.
The profiles of RMSD versus each position at 10 to 66 of Taf1 in 4B0A showed small peaks at 10 to 24 positions and sharp peaks at 40 to 48 positions due to the A140G, R238K and S188L mutations, respectively. As for the P169A mutation, a sharp peak at 40 to 48 positions was also observed (Fig. 7b). Correspondingly, conformational changes at the N-terminus of Taf1-TAND1 were found due to the A140G, R238K and S188L mutations (Fig. 7d). Conformational changes between the Taf1-TAND1 and Taf1-TAND2 regions were observed due to the A140G, P169A, R238K and S188L mutations (Fig. 7d). But for the L214V mutation, no apparent RMSD peak and conformation changes in Taf1 were found (Fig. 7b, d). Remarkably, the A140G mutation seemed to induce the most significant conformational changes of Taf1, implicating that the A140G mutation might have the most significant potential impact on competitive multiprotein TBP interplay patterns [73].